Page MenuHomeFreeBSD

Comitter guide: rewrite the section on testing.
Needs RevisionPublic

Authored by eadler on Jul 2 2018, 1:58 AM.
Tags
None
Referenced Files
Unknown Object (File)
Feb 8 2024, 9:34 AM
Unknown Object (File)
Dec 20 2023, 2:45 AM
Unknown Object (File)
Nov 12 2023, 5:35 PM
Unknown Object (File)
Nov 12 2023, 4:07 AM
Unknown Object (File)
Nov 7 2023, 1:43 AM
Unknown Object (File)
Nov 6 2023, 11:57 PM
Unknown Object (File)
Nov 6 2023, 9:08 PM
Unknown Object (File)
Nov 6 2023, 3:48 AM
Subscribers

Details

Reviewers
emaste
Group Reviewers
Core Team
Summary
  • The previous version felt a little condescending or blame-y. Rewrite

to reduce the blame on the individual but still make strong statements
about needing to test

Diff Detail

Lint
No Lint Coverage
Unit
No Test Coverage
Build Status
Buildable 18386
Build 18101: arc lint + arc unit

Event Timeline

jhb added inline comments.
en_US.ISO8859-1/articles/committers-guide/article.xml
3476–3477

This opening sentence feels too hand-wavy now. While the older one was more snarky, I prefer it to this one. OTOH, maybe something like:

"This may sound obvious, but developers do occasionally break the build. Merging a change tested against an older base revision may break the build or fail to work due to a mismerge. A change may work on one architecture but fail to compile on others, often for seemingly small changes such as &man.printf.3; format strings. Use <command>make tinderbox</command> to compile changes on all supported architectures before committing.

When committing changes to a stable or release branch, always test your changes on a machine
running that branch. Changes to these branches should not introduce build breakages. If merging a change from HEAD which broke the build or introduced regressions, include followup fixes from HEAD in the merge as a single commit.

Refer to the <link ..."

3495

s/resources to help test/testing resources/

I took some of your suggestions. Shortened it a bit.

brooks added inline comments.
en_US.ISO8859-1/articles/committers-guide/article.xml
3479

s/fail/may fail/ ?

3484

fragment?

en_US.ISO8859-1/articles/committers-guide/article.xml
3479

The conditional is at the start of the sentence.

en_US.ISO8859-1/articles/committers-guide/article.xml
3479

Eh, it's not. Here is the sentence in isolation:

"Merging a change against an older base revision fail due to a mismerge."

It definitely needs a 'can' or 'may' before 'fail'.

Also, I do prefer the longer text I had to separate 'build break' from 'fail to work'. I think it's useful to be explicit that this is not just about compile errors but also runtime errors.

3484

I think it's a valid sentence. It's a command / imperative, so the subject is implicit.

This still seems like a useful incremental improvement

I do think it's worth merging, but at the very list the "fail" needs a "may" or "can" in front of it.

emaste requested changes to this revision.Mar 21 2023, 11:18 PM

needs to be refreshed for the asciidoc conversion

This revision now requires changes to proceed.Mar 21 2023, 11:18 PM